PMIx_Fences - remove two during MPI initialization #11451
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
bot:ibm:retest |
hppritcha
force-pushed
the
pmix_fence_removal_v4
branch
from
5fc8716 to
6ea5e84
Compare
This patch removes redundant PMIx Fences in the initialization procedure for MPI when using the World Process Model (WPM). See chapter 11 sections 2 and 3 of the MPI-4 standard for a discussion of the WPM and new Sessions model. The patch does, however, require that what should have been a local operation to support initialization of an MPI session, into a global one. Note this does not disable the sessions feature but just restricts when it will work at this point to use cases that are similar to MPI initialization using the WPM. Refactoring to make ompi_mpi_instance_init_common purely local will require changes that would be too impactive for the current state of the 5.0.0 release cycle. See issue open-mpi#11239. Related to open-mpi#11166 Signed-off-by: Howard Pritchard <[email protected]>
hppritcha
force-pushed
the
pmix_fence_removal_v4
branch
from
6ea5e84 to
c241cb1
Compare
|
@jjhursey could you review this PR when you have a chance? |
|
I feel uncomfortable reviewing this PR. Someone else should review that has time to trace the dependencies. The correctness of MPI_Init over different transports, especially at scale, is sensitive to the structure of the fences. As we saw with the earlier version of the PR, which was later reverted, it is difficult to see those implications through a review - and requires some deeper knowledge of the impact on lower-level transports. Unfortunately, I don't have the cycles to do that level of review. |
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch removes redundant PMIx Fences in the initialization procedure for MPI when using the World Process Model (WPM). See chapter 11 sections 2 and 3 of the MPI-4 standard for a discussion of the WPM and new Sessions model.
The patch does, however, require that what should have been a local operation to support initialization of an MPI session, into a global one. Note this does not disable the sessions feature but just restricts when it will work at this point to use cases that are similar to MPI initialization using the WPM.
Refactoring to make ompi_mpi_instance_init_common purely local will require changes that would be too impactive for the current state of the 5.0.0 release cycle. See issue #11239.
Related to #11166